RX10: Internal AVCHD vs External Recorder.

[Wesley Byram]

Something that has been mentioned by more than one reviewer online, is that the RX10's internal codec is a bit weak. Fortunately, with the clean HDMI output, an external recorder could be a workaround. I didn't see any direct side-by-side comparisons so I decided to actually test this out first hand. To that end, I shot a short, high-detail clip deliberately designed to stress the codec: Handheld, F4, 1/160 shutter, neutral +3,+3,+3. Video was recorded both internally and with a Hyperdeck Shuttle.

These are 1:1 crops of matching frames from each recording:

Internal AVCHD:



Uncompressed via Hyperdeck:


Even without pixel-peeping, the difference is pretty obvious. The interally recorded image has a ton of detail smudged out. Now, bear in mind that this shot was pathologically designed to break the codec, so most of the time, you shouldn't see this much difference.

Of course, we already know that AVC/h.264 footage can break down if you throw too much detail at it, expecially with heavy camera movement. That said, I decided to see how Sony's encoder stands up to another. I took the uncompressed clip and re-encoded it to AVC with x264. Settings there are 23mbps CBR and 1-pass encoding to roughly match the specs of an AVCHD file.

The result:



So, it looks like there's some room for improvement in Sony's implementation here. The x264 encoded file is much less smudgy than the camera's version, even though they are both using h.264 at about 24mbps.

 

[Darren Levine]

Yup, looks good as it should. shame sony won't listen to give us a better codec, my hopes are for a company to make a nanoflash 2 sort of device, an sd/cf fast hd recorder that is as small as possible.

 

[OlegKalyan]

Good to have Ninja2!

 

[tull_power]

Thanks for taking time to make this comparison, it is greatly appreciated. You've added confirmation that the RX10 has a cinema class sensor saddled by a consumer codec. It is not that the AVCHD codec is bad; indeed Sony's implementation of it is the best in the industry. The RX10 just resolves an unprecedented amount of detail which this codec simply cannot keep up with. If the Magic Lantern team or Vitaly (of GH hack fame) would turn their attention to this little gem, I'm sure we'd have an incredible camera at unbelievable bargain. I'm saving up for one to replace my little Nex 5n. Thanks again for the informative test!

 

[bobkrist]

How do we encourage Vitaly or the ML team to do that? I know there is an NEX hack project, but it's kind of in limbo, as far as I can tell. This little camera though, seems to have more potential.

 

[Darren Levine]

there was 'someone' who supposedly got a higher bitrate hack on an nex, but who knows if it's the same architecture as an RX? and as i understood it, the ML hacks were largely possible because of canon first releasing a firmware, something sony likely wont do

 

[Samuel H]

Quite a difference, if only in this pathologically stressful example. Thanks for testing and posting.

 

[Ken7]

Great test, thanks! The moral of the story, if not using an external recorder, is to watch your panning speeds so as to stay within the constraints of the codec. That's a generally good practice anyway.

 

[Wesley Byram]

Indeed, if you intend to use an RX10 to shoot The Blair Witch Project VI: The Sequel No One Asked For, you would do well to invest in a Ninja 2 ;-)

I'll add my voice to the chorus calling for higher in-camera bitrates. On the bright side, at least we're finally starting to see DSLR and "DSLRish" cameras get proper clean HDMI outputs. That's been a standard feature on even low-end consumer camcorders for years and it's always seemed odd to me that DSLR's couldn't do this. I was half expecting to see something amiss in the RX10's HDMI output, hence why I did the test. Fortunately, it appears to work as advertised.

Something that I was a bit concerned about was the reliability of the tiny micro-HDMI port on this camera. I've been using one of these cables to link the RX10 to the Hyperdeck, and it hasn't been a problem so far [knock on wood].

 

[Wesley Byram]

I decided to run one more comparison in order to verify that the HDMI output is indeed 4:2:2 as advertised. I set up a shot of a red and blue color pattern and again recorded the scene with the Hypdereck and internal AVCHD simultaneously. I picked matching frames from each and compaired them. This time the difference was far more subtle.

AVCHD 500% enlargement:



Uncompressed 500% enlarged:


If you look close you'll notice that the AVHD is very slightly softer where the red meets blue. There's a bit more definition in the uncompressed version, especially around the nearly horizontal lines, which is what you would expect of a 4:2:2 signal. Ironically, the AVCHD version looks subjectively better to my eye, as the slight softening hides a subtle stair-step pattern which is likely a hint of aliasing in the image. Edit: Actually, I just looked at my test pattern closely and the stair-stepping is actually in the pattern itself. It's pretty hard to see with the naked eye (especially mine). The hdmi is apparently resolving more real color detail than the AVCHD.

So, it looks like the HDMI output is indeed 4:2:2 as claimed. I expected that it would be, but it never hurts to check.

 

[dan.carter]

I decided to run one more comparison in order to verify that the HDMI output is indeed 4:2:2 as advertised. I set up a shot of a red and blue color pattern and again recorded the scene with the Hypdereck and internal AVCHD simultaneously. I picked matching frames from each and compaired them. This time the difference was far more subtle.

AVCHD 500% enlargement:
View attachment 80530


Uncompressed 500% enlarged:
View attachment 80531

If you look close you'll notice that the AVHD is very slightly softer where the red meets blue. There's a bit more definition in the uncompressed version, especially around the nearly horizontal lines, which is what you would expect of a 4:2:2 signal. Ironically, the AVCHD version looks subjectively better to my eye, as the slight softening hides a subtle stair-step pattern which is likely a hint of aliasing in the image.

So, it looks like the HDMI output is indeed 4:2:2 as claimed. I expected that it would be, but it never hurts to check.

Many thanks for taking the time for these tests and sharing them Wesley. Interesting outcomes.

 

[Ken7]

So Wes, here's the $64 question. Knowing what you now know, seeing what you've now seen, would you make it a practice to an external recorder? If so, would you do this routinely or only for special circumstances?

This probably begs the question are the differences significant enough to warrant the extra effort.

Thanks again for your work on this.

 

[Wesley Byram]

So Wes, here's the $64 question. Knowing what you now know, seeing what you've now seen, would you make it a practice to an external recorder? If so, would you do this routinely or only for special circumstances?

This probably begs the question are the differences significant enough to warrant the extra effort.

Thanks again for your work on this.

That's a good question, and I haven't settled on a firm answer yet. I need to shoot some more to get a better feel for when the codec breaks down. Tentatively, I would say that the AVCHD is good enough for me most of the time. On a tripod shot, AVCHD and uncompressed files look visually indistinguishable, but walking while hand holding the camera can break the codec hard. Such handheld footage wouldn't look that good anyway, AVCHD or not. I plan to play around with flying the camera on a stabilizer this weekend to see if the AVCHD can handle more controlled movement and still be gradeable without noticeable artifacts. Judging by other people's videos on Vimeo, I expect it to do just fine there.

So, most of the time I probably won't bother with the Hyperdeck. Where I would go to the trouble would be:

1. Keying and compositing: Green screen can expose weaknesses in the codec that might otherwise be overlooked, and the 4:2:2 color helps here also. I'll take whatever edge I can get to make my keys come out better. Pretty much any shot that's going to incorporate VFX or otherwise heavy post manipulation is worth being recorded externally.

2. Stuff that's going to festivals: Shorts and music videos that I expect to be played on a theater screen I'll be recording externally for added safety. It's just reassuring to know that I'll be getting the best possible quality out of the camera, and also getting two copies of my footage in case there is a mishap like a card failure. The exception might be steadycam style shots, where the added bulk of the Hyperdeck will likely negate any gains.

 

[Ken7]

Thanks for your thoughtful answer Wes. I look forward to your further evaluations.

BTW, I guess you're OK with the absence of 1920x1080 60p recording on the Hyperdeck?

 

[Wesley Byram]

Thanks for your thoughtful answer Wes. I look forward to your further evaluations.

No problem. Speaking of further evaluations, I did do a few tests shots with the RX10 on the Flycam, sans Hyperdeck. Nothing fancy, I just flew the camera around the yard for a few minutes, then did some minor corrections in post: slightly lifted the shadows, nudged up the saturation just a pinch, and added a tiny bit of sharpening. Slow, floating camera movements hold up just fine, but there is visible macroblocking on sudden movements, like whip pans. Fast pans can sometimes look pixelated, instead of smoothly blurred like they should. That could probably be smoothed over in post, but it's something to note.

I feel pretty confident in saying that the internal codec cannot be trusted for action heavy shots. If I were going for a Saving Private Ryan sort of look with this cam, I would definitely record externally.

The way I look at it, most cameras in this price range need some accessories to really get everything out of them. The BMPCC for example needs batteries, an EVF or a least a loupe, separate sound recorder, and a rig to lash it all together. That's not even counting lenses, of course. IMHO the RX10 really just maybe needs an HDMI recorder, and some kind of simple rig or cage to mount it.

That's not so bad. For around two grand you can get an RX10 and a Hyperdeck/Ninja and have a very potent combo that can produce some really great looking, minimally compressed 4:2:2 footage. IMHO it doesn't need an add-on monitor or EVF, as the ones built-in are quite good, even in sunshine, and you get zebras and peaking. You don't have to worry about lenses, because it's got a good one of those built-in. (On the flip side, it's the ONLY lens you've got on this camera, so that could be an issue for many.) It also takes pretty respectable stills.

BTW, I guess you're OK with the absence of 1920x1080 60p recording on the Hyperdeck?

Well, it is what it is. It would be nice if it could record 1080p60, but the Hyperdeck is cheap and I don't expect miracles. 1080p60 is twice the data rate of 1080i60 and most SSDs have trouble just keeping up with that. Besides, I don't use 60p all that much. 24p is fine 99% of the time.

Now, one thing that I DO really hate about the Hyperdeck is that it doesn't remove pulldown. Truth be told, I sometimes wish that I spent the few extra bucks for a Ninja 2 for that reason alone. Come on Blackmagic, give us a firmware update or something.

 

[toxotis70]

is there any external recorder with 50-60p recording ?

 

[Propertv.be]

is there a time limit for using an external recorder? I'm looking for a cheap multicam camera, but i'm afraid it will enter sleepmode after some time.

 

[Wesley Byram]

is there any external recorder with 50-60p recording ?

I believe that the Odyssey7 can. At $1300, it's a bit spendy though.

 

[Wesley Byram]

is there a time limit for using an external recorder? I'm looking for a cheap multicam camera, but i'm afraid it will enter sleepmode after some time.

By default, the camera does go to sleep after a minute. You can change that in the menus though. It can go as high as 30 minutes before sleep activates.

 

[szimmer]

I was looking at the Odyssey 7 last night. It seems to not be out yet until March 2014. I also read it only works with the SSD's from Convergent Design. They are pricey at ~$400 for 250GB.

I am about 99% certain I will get the Sony RX this week. Still debating whether it is worth it to get Ninja 2 or wait for Odyssey 7.

I do videos for our water non-profit organization in Colombia. I need a run and gun and use Avid Media Composer. Therefore the Ninja 2 sounds great except it can not capture 1080 60p. My final output will be for vimeo or presentations from my laptop. Will the quality of 1080 60p be so much better than the internal codec?

Simon